Sophos data scientist: AI harm forecasting is far harder than Covid's was
joshua_saxe · x · 2026-08-27
Security researcher Joshua Saxe says he has grown uncomfortable with Covid analogies for AI harms: pandemics offer a statistically meaningful reference class and transmission depended less on social complexity, so harms could be modeled smoothly — AI lacks both properties, making harm prediction intrinsically harder.
More from AGI Musings
- OpenAI and Anthropic take a third of incremental world compute, Dylan Patel says — bookwormengr · 2026-08-27
- Do LLMs have emotions? Distinguishing functional from sensory — VoidStateKate · 2026-08-27
- Ex-Google researcher Szegedy: denying AI's transformation of math is "delusional" — ChrSzegedy · 2026-08-27
- Users Less Likely to Critique AI in High-Stakes Tasks — EchoShao8899 · 2026-08-27
- Apple's New Ad Proves 'Show Your Work' Is the New Value in the Age of AI — adariostrange · 2026-08-27
- Opinion: We are still largely in the command-line era of AI — BrandNewFeel · 2026-08-27